Environmental controls on the spatial and temporal variability of savanna productivity in the Northern Territory, Australia

The overall objective of this thesis is to examine the role of environmental drivers in controlling the spatial and temporal variations of gross primary productivity (GPP) in savannas in the Northern Territory (NT), Australia. GPP is a critical measure of the health and sustainability of ecosystems. An understanding of GPP will underpin predictions of the impact of climate change on the savanna carbon cycle. This thesis employs an integrated approach combining in situ measurements, eddy covariance based flux tower data and remote sensing techniques to address the objective. The control of aerosols and clouds (important environmental variables in savannas) on the temporal variation of GPP was examined at Howard Springs, a tropical savanna site in the NT. Results indicated that aerosols and clouds affected the temporal variability of savanna productivity by altering the quantity and quality (partitioning of total radiation into direct and diffuse) of solar radiation. It was found that in the dry season aerosols emitted in the region were relatively lower compared to other savanna regions. Consequently, a small increase in the diffuse radiation (22%) resulted in a small decrease in the total radiation (10%), but did not affect the GPP significantly. In contrast, during the wet season, diffuse radiation increased due to cloudiness but was overshadowed by large decreases in total radiation (57% under thick clouds compared to clear sky periods) which reduced overall productivity by up to 19% under thick clouds. Information on other environmental controls such as fPAR (fraction of absorbed Photosynthetically Active Radiation), VPD (vapour pressure deficit), soil moisture, and temperature on the temporal variability of savanna GPP were also examined using Moderate Resolution Imaging Spectro- radiometer (MODIS) GPP products. Given that these products are generated based on inference from surface reflectance, MODIS GPP and the upstream products used to estimate GPP; fP AR, light use efficiency (LUE) and climate were validated against flux tower derived GPP to improve the products for savannas. In northern Australia, soil moisture rather than VPD was found to be an important factor limiting savanna GPP in the dry season. This was suggested by the improved estimation of GPP and improved simulation of the seasonal dynamics when VPD was replaced with a moisture index in the GPP algorithm. Validation of MODIS GPP and associated inputs provided information on the uncertainties of the MODIS GPP algorithm inputs. This information was used to improve the estimation of GPP at the regional scale across the NT from 2000 to 2007 using field based LUE, regional specific meteorology and fPAR from the latest MODIS product. Results showed that GPP estimated with this approach captured the seasonal patterns of monthly GPP quite well across 18 sites along the Northern Australian Tropical Transect (NATI). The magnitude of GPP was also estimated quite well with only a 6% error at the Howard Springs site. Changes in rainfall along the gradient resulted in changes in the structure and composition of savannas and GPP across the NA IT. Consequently, within the NT savanna region, vegetation type was a major driver of GPP with closed forest having six times more GPP than Acacia vegetation. Examination of the environmental controls on the spatial variation in GPP showed a strong influence of mean annual rainfall (r2 0.88). In terms of inter- annual variability, arid ecosystems had higher variation (>20%) in GPP than forests (<10%) and this was associated with large variations in rainfall (>30% for arid vegetation versus 19% for forest). These results suggest that future changes in precipitation driven by climate change may affect the future distribution and dynamics of savanna vegetation in northern Australia. Given the significance of soil moisture (rainfall) and radiation control on GPP over Australian savannas, more emphasis should be placed on models that are able to accurately simulate the sensitivity of these factors on productivity. Results presented in this thesis provide valuable information for savanna management in predicting the response of savannas to perturbations of the major environmental drivers. Such research will become vital in formulating strategies to secure resources in the tropical savannas of northern Australia, as well as mitigating the potential adverse effects of climate change.